MCP Client with React Flow and ollama

[mario-andreschak]

Pre-submission Checklist

• This post follows the MCP community guidelines

What would you like to share?

I wanted to showcase the MCP Client FLUJO.

FLUJO is a neat all-in-one application for managing AI models, MCP servers, and complex workflows—open-source and local-first.

Built with Cline and PocketFlowFramework, it's a great example of what the MCP community is creating.

2/ Key features:

• Visual Flow Builder for AI workflows
• Model management across different providers
• MCP server integration
• Secure API key storage
• Chat interface with file attachments
• OpenAI-compatible endpoint

3/ The Flow Builder is pretty slick. FLUJO allows to design multi-step AI workflows visually, connecting different models together. You can configure each node with specific prompts and tool access without diving into code.

4/ For model management, we set things up so you can configure multiple AI models simultaneously—OpenAI, Anthropic, or local Ollama models. Nice touch adding custom system instructions for each model.

5/ The MCP integration shows how FLUJO is leveraging Model Context Protocol. FLUJO makes it straightforward to install MCP servers from GitHub, check out available tools, and connect them to workflows.

How to install, troubleshoot & test a MCP Server in FLUJO in ~2 mins

6/ The chat interface is practical—letting you attach documents or audio files, manage conversation context, and let your workflow handle the heavy lifting behind the scenes.

7/ Versatility is key—FLUJO works as both a web app and a desktop application. Run it locally or in the future: deploy it on edge devices depending on what you need.

8/ If you want to check out what we built:

1. Download from GitHub or clone the repo
2. Run with npm install followed by npm start
3. Access at http://localhost:4200
4. Set up your models, MCP servers, and start building

9/ This tool could be useful for developers, AI engineers, or researchers who want to create custom AI workflows without cloud platform lock-in. Good for experimentation or building specialized tools.

10/ FLUJO is still in early preview. If you're interested, check it out on GitHub:

https://mario-andreschak.github.io/FLUJO/githubpages/index.html

Relevant Links

https://github.com/mario-andreschak/FLUJO/